Objectives: To examine the dimensionality of scales for measuring patient-reported cognitive and communicative functioning in a sample of stroke survivors and to evaluate the consequences for measurement of treating them as a single, undifferentiated construct.
Design: Secondary analysis of existing cross-sectional data.
Setting: Data were collected in outpatient rehabilitation clinics and in the community.
Participants: Unilateral stroke survivors (N=316) 3 months or more postonset referred for participation in research.
Interventions: Not applicable.
Main outcome measures: The Burden of Stroke Scale cognition and communication domain scales were evaluated by using confirmatory factor analysis, Rasch analysis, and tests of differential item functioning (DIF). To evaluate the impact of multidimensionality on the measurement of individual patients, separately estimated cognition and communication scores were compared. Combined and separately estimated scores were also examined for responsiveness to group differences in the presence of cognitive and communicative impairment.
Results: Factor analysis and Rasch model fit analyses equivocally supported the unidimensionality of the item pool. DIF analyses between participants with right versus left hemisphere stroke suggested multidimensionality. Scaling cognition and communication items separately resulted in different person scores for a significant number of patients and greater responsiveness to group differences.
Conclusions: Patient-reported scales assessing communication along with more general cognitive activities may possess an internal structure that is inconsistent with a unidimensional measurement model with potential negative consequences for measurement.
Published by Elsevier Inc.